nitrogen-induced changes in seedling regeneration …kobe/docs/catovsky et al 2002 ecol appl.pdf ·...

1611

Ecological Applications, 12(6), 2002, pp. 1611–1625q 2002 by the Ecological Society of America

NITROGEN-INDUCED CHANGES IN SEEDLING REGENERATION ANDDYNAMICS OF MIXED CONIFER–BROAD-LEAVED FORESTS

S. CATOVSKY,1,3 R. K. KOBE,2,4 AND F. A. BAZZAZ1

1Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, Massachusetts 02138 USA2Department of Forestry, Michigan State University, East Lansing, Michigan 48824 USA

Abstract. Most research on forest dynamics has focused on species’ light requirementsas the major driver for successional change. However, soil resource availability may modifyseedling responses to light and ultimately alter the course of succession. In the presentstudy, we examined how seedlings in mixed conifer–broad-leaved forests in eastern NorthAmerica differed in their growth and mortality responses to manipulated nitrogen avail-ability. We then incorporated these responses into an individual-based model of forestdynamics (SORTIE) to assess potential longer-term consequences of seedling responses tonitrogen for temperate forest community dynamics. We grew seedlings of six study species,both individually and in mixed-species competitive stands, in a common garden for twoyears. The earlier successional broad-leaved species (yellow birch and red maple) consis-tently showed the greatest increases in biomass in response to nitrogen addition, while themost late successional of the broad-leaved species (sugar maple) and all the coniferousspecies did not grow significantly larger with increased nitrogen. We found a significantcorrelation between species’ early growth rate and nitrogen growth enhancement. For thosespecies that underwent significant nitrogen-induced shifts in growth and/or mortality, weadjusted their parameters in the seedling/sapling growth and mortality submodels of SOR-TIE (covering up to 10 cm dbh). Simulations revealed that nitrogen effects on both seedlinggrowth in high light and seedling mortality in low light (data from parallel experiment)changed overall forest structure and dynamics. Increased nitrogen led to: (1) further dom-inance of young forests by earlier successional species (yellow birch in particular), throughits impacts on seedling high-light growth, and (2) even greater persistence of later suc-cessional species (predominantly hemlock) in older forests, through its impacts on seedlinglow-light mortality. These findings were robust to an uncertainty analysis that incorporatedexperimentally derived error into the seedling/sapling submodels. In contrast, the identityof the species replaced by yellow birch and hemlock was more sensitive to uncertainty inparameter values. We conclude that seedling physiological and demographic responses toincreased nitrogen availability have the potential to scale up and influence successionaldynamics in mixed temperate forests, provided these effects persist throughout seedlingand sapling life stages.

Key words: forest gap; forest succession; mixed conifer–broad-leaved forests; nitrogen avail-ability; seedling growth; seedling regeneration; SORTIE model; temperate forest community dynamics.

INTRODUCTION

Forests are now recognized as dynamic communities,characterized by change and disturbance rather thanstasis (Foster et al. 1996). As a result, there has beena growing realization that a mechanistic understandingof forest structure and composition requires the inves-tigation of processes that regulate the population dy-namics of individual component species (Shugart1998). To date, most research has focused on tree spe-cies’ effects on and responses to light availability asthe major drivers of forest successional dynamics (Can-ham et al. 1994, Pacala et al. 1994, Kobe et al. 1995).

Manuscript received 11 May 2001; revised 8 February 2002;accepted 25 March 2002.

3 Present Address: NERC Centre for Population Biology,Imperial College at Silwood Park, Ascot, Berks SL5 7PYUK.

4 Corresponding author. E-mail: [email protected]

However, soil resource availability could modify seed-ling responses to light, and ultimately alter the courseof forest succession (Kobe 1996, Finzi and Canham2000, Caspersen and Kobe 2001). Understanding ni-trogen effects in temperate forests is now especiallycritical because: (1) nitrogen is a particularly limitingsoil resource in these forests (Vitousek and Howarth1991), and (2) human activities are leading to increasednitrogen loading in such systems (Vitousek et al.1997a).

To assess the potential effects of nitrogen availabilityon temperate forest community structure, we need toexamine how nitrogen influences the different pro-cesses controlling successional trajectories in these for-ests. Species growth and mortality early in life haveboth been identified as particularly important driversof forest dynamics (Pacala et al. 1996). However, wecurrently have relatively limited information on how

1612 S. CATOVSKY ET AL. Ecological ApplicationsVol. 12, No. 6

nitrogen availability affects regeneration patterns, andhow these nitrogen effects might scale up to influenceforest dynamics in the longer term, i.e., over the courseof forest succession. Part of the difficulty arises becauseprevious research has examined nitrogen effects on justone aspect of regeneration, e.g., growth in forest gapsvs. the understory (Canham et al. 1996, Walters andReich 1997), and done so without direct manipulationof nitrogen in the field (reviewed in Grubb 1994). Thus,to provide a more comprehensive view of nitrogen ef-fects on forest dynamics, we designed two parallel ex-periments to examine seedling responses to nitrogenaddition at two critical stages in a tree’s life cycle: (1)persistence in the understory seedling bank (low light)(Marks and Gardescu 1998) and (2) seedling growthafter canopy gap formation (high light) (Canham 1989).Full results of the understory study are reported else-where (Catovsky and Bazzaz 2002a).

Mixed temperate forests may be particularly sensi-tive to variation in soil nitrogen availability becausethey contain two very different groups of tree species:evergreen coniferous and deciduous broad-leaved.These groups may respond differently to increasingnitrogen, given their differences in leaf longevity andlife-history strategy (Bond 1989, Becker 2000). In thepresent study, we examined how seedlings of three co-niferous and three broad-leaved species common tomixed temperate forests in eastern North America dif-fered in their responses to nitrogen addition. We testedthe prediction that coniferous and late-successionalspecies would be less responsive to changes in nitrogenavailability than would broad-leaved and early-succes-sional species, as their evergreen habit and slow seed-ling growth rates often correlate with more conserva-tive patterns of nutrient uptake and use (Reich et al.1995, 1998b, Bazzaz 1996, Aerts and Chapin 2000).Because seedlings typically regenerate in dense standswithin forest gaps (Peet and Christensen 1987), weconsidered responses of seedlings grown both individ-ually and in mixed-species competitive stands. We thenexamined how the seedling growth and mortality re-sponses to increased nitrogen availability influencedforest successional dynamics using SORTIE, a spatiallyexplicit, individual-based model of forest dynamics(Pacala et al. 1996). Rather than precisely predictingsuccessional trajectories under different nitrogen re-gimes, we used SORTIE as a tool to highlight the po-tential consequences of seedling responses to soil ni-trogen for temperate forest dynamics.

MATERIALS AND METHODS

Soil and seed collection

Soil was collected from hemlock- and red oak-dom-inated forest stands at Harvard Forest (Petersham, Mas-sachusetts, USA; 428329 N, 728119 W, elevation 340m) in April 1998. The stands were located in the TomSwamp tract and were chosen so that hemlock and red

oak contributed .50% of the basal area in each of threestands. Details of the stands are described elsewhere(Catovsky and Bazzaz, in press). The soil was collectedfrom one area in each stand as intact soil divots (40 350 cm, 15 cm depth to include all of O and some ofA horizon), and was then transported back to an ex-perimental garden at Harvard University (Cambridge,Massachusetts, USA). Using a trowel, divots were di-vided vertically into smaller soil squares, which wereplaced over coarse silica sand in plastic containers.Individuals were grown in 12-L tree pots (16 3 16 cm,48 cm deep), with ;8 L of sand beneath a 16 3 16 315 cm soil square. For mixed-species plantings, weused wider but shallower containers (20 L; 30 3 35cm, 20 cm deep), with a 5-cm layer of sand at thebottom. This potentially increased root as well as shootcompetition, and led to an overall reduction in speciesallocation to roots when grown in stands vs. individ-ually (36% vs. 53%). We initially planned to look atdifferences in species’ responses between stand typesas well, but subsequent analyses revealed that the soildid not differ significantly between hemlock and redoak stands (F1,93 5 3.00 for ammonium, 0.35 for nitrate,P . 0.05 for both), and neither did species’ responses(F1, 232 5 1.74 for individually grown seedlings, 0.04for competitively grown seedlings, P . 0.05 for both).Thus, for the purposes of this experiment, species’ re-sponses were pooled across stand types. However, wekept a ‘‘site’’ term in our models, so that variation fromour six different study locations could be taken intoaccount.

We chose tree species that spanned a range of shade-tolerance classes (Table 1) from both gymnosperm andangiosperm groups, enabling us to distinguish differ-ences in species’ responses due to evolutionary historyfrom life-history strategy. In addition, all species arecurrently important components of mixed temperateforests in New England (Foster et al. 1998). Seeds ofall study species were collected from multiple trees atHarvard Forest in the autumn of 1996 (most species)and spring of 1997 (red maple). Seeds were air-driedand stored at 48C until late autumn 1997, when theywere placed in cloth bags and buried in trays of wet,coarse sand. These trays were placed outside throughthe winter to stratify the seeds and were collected thefollowing spring. In 10 cm deep germination flats,seeds were spread out evenly over a peat-based pottingmix with added perlite, and then covered with a thinlayer of vermiculite. Flats were placed in the experi-mental garden in mid-April and seeds were left to ger-minate. The flats were monitored daily and wateredwhen necessary. Seedlings began to germinate in earlyto mid-May, and were transplanted into containers inmid-June when most seedlings had two to three trueleaves. We planted seedlings in both ‘‘noncompetition’’and ‘‘competitive mixed-species’’ treatments. For thenoncompetition treatment, one seedling was plantedper individual pot. For the competition treatment, 42

December 2002 1613NITROGEN AND DYNAMICS OF MIXED FORESTS

TABLE 1. Details of experimental study species in hemlock- and red oak-dominated foreststands at Harvard Forest, Petersham, Massachusetts.

Species Common name Leaf habitSuccessional

position†

Tsuga canadensis (L.) Carr.Picea rubens Sarg.Pinus strobus L.Acer saccharum Marsh.Acer rubrum L.Betula alleghaniensis Britt.

eastern hemlockred sprucewhite pinesugar maplered mapleyellow birch

evergreen coniferousevergreen coniferousevergreen coniferousdeciduous broad-leaveddeciduous broad-leaveddeciduous broad-leaved

543543

† Based on Baker’s Table and mortality–light relationships in Kobe et al (1995). Successionalposition/shade tolerance is on a scale of 1–5, where 1 is the least and 5 is the most shadetolerant.

seedlings were planted in each container, representinga seedling density of 390 individuals/m2, approximat-ing the mean density (300 individuals/m2, range: 30–1200 individuals/m2) for seedlings established from theseed bank following a canopy disturbance in this region(Catovsky and Bazzaz 2000). In the mixed-speciestreatments, seedlings were planted in a hexagonal ar-ray, such that each ‘‘target’’ plant had six neighbors.Three individuals of each species were planted in the18 central ‘‘target’’ locations, and four individuals ofeach species were planted in the 24 outer ‘‘edge’’ lo-cations. Seedling positions were randomized withinthese target and edge locations, and the positions wereestablished using a planting template with holesmarked for each seedling.

Experimental treatments and growth conditions

We grew 54 individuals of each species and 54mixed-species plantings for the experiment: 6 sites 33 nitrogen levels 3 3 replicates. All containers for in-dividuals and mixed plantings were buried in theground in the experimental garden underneath a largeplastic greenhouse (30 m long, 6.5 m wide, 3.5 m tall),in which a continuous flow of air was maintained withthe use of large embedded fans at one end. The plasticcovering of the greenhouse resulted in 60% of possiblephotosynthetically active radiation and was chosen tosimulate light levels in a multiple tree-fall gap in tem-perate forests (Bazzaz and Wayne 1994). These largerscale disturbances dominate the major successionalchanges that occur within New England forests (Oliverand Stephens 1977, Peterken 1996), producing a dis-turbance intensity of 1% annually (when averaged overtime) (Canham and Loucks 1984). Within the green-house, the individuals and mixed-species plantingswere arranged in three large blocks, with one replicateof each treatment combination placed in each block.Seedlings were given daily watering treatments fromautomatic sprinklers within the greenhouse. Water wasapplied generously so that it would not be a limitingfactor in this experiment.

Each individually grown seedling or mixed plantingreceived one of three nitrogen addition treatments (0,

2.5, or 7.5 g N·m22·yr21). In both 1998 and 1999, seed-lings were given nitrogen eight times per year at;three-week intervals, beginning mid-April and end-ing mid-September. At each addition, nitrogen was ap-plied as dissolved ammonium nitrate solution of vary-ing concentration. Individual seedlings were given 10mL of 0, 0.030, and 0.090 mol N/L, and mixed plant-ings were given 100 mL of 0, 0.012, and 0.036 mol N/L. The treatments were designed to provide a substan-tial perturbation to the natural nitrogen cycle, repre-senting nitrogen treatments that added considerably tonet nitrogen mineralization (4–8 g N·m22·yr21) and ni-trification rates (0.5–2 g N·m22·yr21) (Catovsky andBazzaz, in press), and current levels of nitrogen de-position (0.6 g N·m22·yr21) (Munger et al. 1998) atHarvard Forest. Nitrogen was added as ammonium ni-trate, because both ammonium and nitrate each com-pose close to half of the deposition in New England(Ollinger et al. 1993).

Seedling mortality censuses were taken at the end ofthe first growing season (1998), and at both the begin-ning and end of the second growing season (1999).Fallen litter of deciduous species was removed betweenthe two growing seasons. In late September 1999, allseedlings were harvested (16 mo after start of exper-iment). Leaves, stems, and roots of all seedlings wereseparated. Isolated roots were obtained by carefullywashing away the soil. For mixed-species stands, everyindividual’s root system was carefully separated byhand. Plant material was dried at 708C for 7 d and thenweighed.

Multifactor analyses of variance were used to in-vestigate influence of nitrogen addition on seedlinggrowth and mortality (Sokal and Rohlf 1995). Linearmodels included nitrogen addition as a continuous fac-tor, species as fixed discrete factors, and site (for thesix different study locations) and block within green-house both as random factors. The mixed-species plant-ing analysis also included a nested plot term repre-senting each planting. Significant multifactor interac-tions involving nitrogen were investigated by exam-ining the magnitude and significance of regressionslopes (dependent variable vs. nitrogen addition). Bio-


mass data were natural logarithm transformed to ensurethat the assumptions of analysis of variance were met(normality of residuals, homoscedascity).

Seedling biomass and nitrogen measurements

In August 1999, a leaf sample was taken from everysurviving seedling growing individually. Samples werephotocopied and leaf area subsequently calculated us-ing NIH Image software version 1.6 (NIH, Bethesda,Maryland, USA). The leaf samples were weighed afterdrying in an oven for 48 h at 708C, and specific leafmass calculated (SLM, g/m2). Dried leaf material wasground using a Mikro-Dismembrator (B. Braun BiotechInternational, Allentown, Pennsylvania, USA). A sub-sample (2–5 mg) of the ground leaf was used to de-termine foliar nitrogen concentrations, using a FisonsCHN Analyzer 1500 Series 2 (Beverly, Massachusetts,USA).

Season-long integrated nitrogen availability was de-termined using ion exchange resin bags placed in allmixed-species plantings and in a subset (one-sixth) ofindividually grown seedlings from June until Septem-ber 1999 (Binkley and Vitousek 1989). The bags wereconstructed with 22 mL of mixed bed strong acid (cat-ion) and strong base (anion) gel resins (Sybron Chem-icals, Birmingham, New Jersey, USA) sealed in nylonmesh, and placed at a depth of 5 cm in the soil. Afterremoval from the soil, 4 g of dried resin (708C, over-night) was extracted with 100 mL of 2 mol/L potassiumchloride solution (258C, 24 h), and then frozen im-mediately following suction-filtration. Ammonium andnitrate in all soil and resin extracts were measured usinga Lachat continuous flow ion analyzer using methods12-107-06-1-A and 12-407-04-1-B (Lachat Instru-ments, Milwaukee, Wisconsin, USA). Blanks were cre-ated from resin bags that had been sealed in polyeth-ylene bags for the length of the growing season. Theseresins were extracted in the same way as the resinsplaced in the soil and were used to determine the lowerthreshold of detection.

SORTIE parameterization

To explore changes in community dynamics medi-ated through seedling responses to nitrogen availabil-ity, we incorporated results of the nitrogen fertilizationexperiments into the forest dynamics model SORTIE.We used the model as a tool to explore how nitrogen-induced changes in seedling demography could poten-tially alter the dynamics of mixed temperate forests.The modeling exercise was not designed to generatespecific predictions about effects of nitrogen avail-ability on successional trajectories in these mixed for-ests, but rather to highlight the potential importance ofseedling responses to soil nitrogen in driving these dy-namics.

We summarize the important attributes of SORTIEhere to provide a context for our nitrogen parameter-izations, but see Pacala et al. (1996) for a detailed

description and analysis of the model. SORTIE predictslong-term community changes in tree species density,spatial distributions, and age and size structure by re-peated iterations of four submodels governing individ-ual tree behavior: (1) seedling recruitment as functionof distance to and size of parent trees, (2) light ex-tinction through crown interception, (3) seedling/sap-ling growth as a function of light availability, and (4)seedling/sapling mortality as a function of recentgrowth. The four submodels of SORTIE were cali-brated from field studies for nine major tree species incentral/southern New England. SORTIE operates on afive-year time step and tracks every individual tree incontinuous space.

A model run begins with user-specified numbers, siz-es, and species of individual trees. Subsequent estab-lishment of seedlings arises from the reproduction oftrees on the simulated plot, governed by a species-specific seedling recruitment submodel (Ribbens et al.1994). The submodel specifies the number of seedlingsproduced by an individual tree, which is scaled to treediameter, and the spatial distribution of those seedlings.Each tree in SORTIE experiences a light environmentas modified by neighbors (Canham et al. 1994). In-coming radiation, characterized by a spatial distribu-tion of sky brightness, is attenuated as it interceptscrowns of individual trees. Tree crowns are representedin the model as cylinders, whose dimensions are de-termined by species-specific allometric equations. Be-sides crown volume, species also differ in leaf andbranch density. Thus, the amount of light reaching aparticular tree is calculated as the attenuation of lightfrom the sky hemisphere, passing through reconstruct-ed crowns with species-specific light extinction. Het-erogeneity in light availability is caused by variationin the size, density, and composition of neighbors.Light availability in turn determines the species-spe-cific growth rate of juvenile trees (#750 cm in height)as a Michaelis-Menten function (Pacala et al. 1994):

G 3 light1Dradius 5 radius 3 (1)G1 1 lightG2

where G1 (the asymptote) and G2 (slope of the growthfunction at zero light) are estimated from field data andgovern high-light and low-light growth, respectively.Trees .750 cm height add a constant area incrementfor each time step of the model. Juvenile trees die witha probability that is a function of recent growth rates(Kobe et al. 1995). The relationship between proba-bility of mortality in any given 2.5-yr period and thepreceding five years of radial growth is characterizedby an exponential function:

2M 3g2P(mort) 5 M 3 e2.5 yr 1 (2)

where g is the average radial growth rate over the mostrecent five years and M1 and M2 are species-specific


parameters estimated from data with 0 , M1 # 1. M1

represents mortality at zero growth (typically at verylow light levels), and M2 characterizes the sensitivityof mortality to changes in carbon balance or recentgrowth rates. Non-juvenile trees experience a constantprobability of mortality (Pacala et al. 1996).

Here we focused on the potential effects of increasednitrogen availability on community dynamics as me-diated by the seedling growth and mortality responsesto nitrogen of the six study tree species. Although SOR-TIE has not been calibrated for Harvard Forest, ourinterest was not to simulate this site per se, but to assessif the magnitude of the observed seedling growth andmortality responses to increased nitrogen had the po-tential to alter species composition and successionaltrajectories of such mixed temperate forests. We ranthree sets of SORTIE simulations: control (0 g N·m22·yr21), low nitrogen addition (2.5 g N·m22·yr21), andhigh nitrogen addition (7.5 g N·m22·yr21). For the SOR-TIE control simulations, we used model parameters asreported for Great Mountain Forest (GMF) in northwestConnecticut. GMF and Harvard Forest are similar insoils, species composition, and climate, and are broadlyrepresentative of central New England forests (West-veld 1956). With the exception of red spruce, all speciesin our nitrogen fertilization experiments were includedin the GMF calibration of SORTIE. Red spruce showssimilar growth, mortality, and foliage density as easternhemlock (Burns and Honkala 1990), so we substitutedhemlock parameter values for red spruce in these sim-ulations.

For the increased nitrogen simulations, we adjustedthe parameters of the seedling/sapling growth and mor-tality submodels of SORTIE relative to the control forspecies that underwent significant shifts in these pa-rameters following nitrogen addition. These submodelscover seedling/sapling performance up to 10 cm di-ameter breast height (dbh), after which the adult re-sponse functions take over.

Growth.—Nitrogen additions only influenced growthunder the high light conditions of a simulated gap (currentexperiment), but not in the low light of the understory(Catovsky and Bazzaz 2002a). Thus, to simulate nitrogeneffects on growth of each species, we multiplied the highlight growth parameter (G1) in Eq. 1 by the ratio of av-erage seedling growth in each nitrogen treatment to av-erage seedling growth in the control (GrowthN/GrowthC)to obtain estimates of G1,N for both the low and highnitrogen amendments. If nitrogen additions did not resultin a significant change in growth, the enhancement ratiowas kept at unity. We only used growth responses fromseedlings grown without competition, because thosegrown with competition experienced a variable light en-vironment, depending on their position within the canopy.As SORTIE measures growth as changes in radius ratherthan biomass, we converted the relative biomass changes(GrowthN/GrowthC) to relative diameter growth changesusing allometric growth equations of the form M 5 aDb,

where M is biomass, D is stem diameter, and a and b areallometric constants (ter Mikaelian and Korzukhin 1997).

Mortality.—Nitrogen additions only had significanteffects on seedling mortality under the low light con-ditions of the understory experiment (Catovsky andBazzaz 2002a), but not in high light gap conditions(current experiment). To simulate these nitrogen ef-fects, we modified mortality submodel parameters (M1

and M2, Eq. 2) according to nitrogen-induced changesin mortality relative to the controls (%MortalityN/%MortalityC). Species’ parameters were adjusted whennitrogen significantly altered mortality in at least oneof the experimental treatments used in the understoryexperiment: planted seedlings, seed addition, or naturalregeneration (Catovsky and Bazzaz 2002a). If nitrogensignificantly affected more than one of these treat-ments, then the pooled average across these treatmentswas used. In SORTIE, average mortality for a speciesunder a particular nitrogen condition can be calculatedby integrating Eq. 2 for positive growth rates. Thiscalculation indicates that average mortality can be sum-marized as M1/M2. Thus, nitrogen-induced changes inmortality for both the low and high nitrogen treatmentsare characterized by multiplying M1 or dividing M2 by%MortalityN/%MortalityC to obtain M1,N or M2,N esti-mates for the low and high nitrogen treatments. Formost species, we modified M1, the probability of mor-tality at zero growth (or very low light) (Eq. 2), becauseit most closely represented the mortality data from theexperiment (carried out under the very low light con-ditions of the understory). The only exception was redmaple, whose M1 estimate is 0.99 (Kobe et al. 1995).Because M1 represents the probability of mortality atzero growth and because red maple showed increasedmortality with nitrogen addition, it was biologicallynon-sensible to alter M1 to a value greater than 1. In-stead, we modified red maple’s M2 parameter to sim-ulate the nitrogen-induced increase in mortality. Fur-thermore, in SORTIE, species-specific mortality de-pends on an individual’s recent radial growth. To un-couple the independent effects of nitrogen additions ongrowth and mortality, mortality was corrected for ni-trogen effects on growth by dividing M2 (decay param-eter determining how quickly mortality decreases withincreases in growth) by the growth enhancement undera particular treatment.

We modified species’ parameter estimates for theSORTIE growth and mortality submodels as describedabove in order to assess potential consequences of twolevels of increased nitrogen availability on forest com-munity dynamics. All other parameters in these sim-ulations were the same as in the control simulations,i.e., all parameters controlling seedling recruitment,tree allometric relationships, mature tree growth andmortality, and attenuation of light availability. We ran40 replicate 9-ha simulations with different randomnumber seeds for each of the control, low nitrogen, andhigh nitrogen treatments (referred to as ‘‘baseline


runs’’). We present mean relative basal areas for theseruns and the range for the central 95% of runs, whichrepresent stochastic variation in model predictions. In-herent stochasticity in the model arises from the twoprobabilistic submodels; one specifying seedling/sap-ling mortality probability and the other the probabilityof seedling establishment.

SORTIE uncertainty analysis

We also performed an uncertainty analysis to assesshow error in species’ mean responses to nitrogen ad-ditions translated into uncertainty in SORTIE com-munity-level predictions. We calculated variance in ra-dial growth enhancements (GrowthN/GrowthC) usingthe following estimator for variance of a ratio for in-dependent samples (de Vries 1986):

2Growth Growth (var Growth )N N Nvar 5 321 2 1 2Growth Growth (Growth )C C Ni

(var Growth )C3 . (3)2(Growth )C

The ratio variance, normalized by the smaller samplesize of the nitrogen or control treatments, provided anestimate of the standard error of the growth enhance-ment for each species under low or high nitrogen ad-ditions relative to controls. A sampling distribution foreach species-treatment growth enhancement was de-veloped as a normal distribution centered about themean growth enhancement, with the standard error ofthe mean growth enhancement.

The sampling distribution represents the distributionof mean growth enhancements under a nitrogen treat-ment relative to the control. From the sampling distri-bution, we generated 40 possible mean growth en-hancements for each species (6) and nitrogen treat-ments (2) using S-Plus 2000 (MathSoft Corporation,Seattle, Washington, USA). In a similar manner, wegenerated 40 possible mean mortality alterations foreach species and nitrogen level. For each nitrogen treat-ment, 40 mean growth and mortality manifestationswere compiled across species to result in 40 sets ofspecies-specific growth and mortality changes, whichwere then used to alter SORTIE parameters as de-scribed above. For each nitrogen treatment, this re-sulted in 40 SORTIE parameter sets, encompassing thestatistical uncertainty that we observed in species-spe-cific mean seedling growth and mortality responses tonitrogen additions. The 40 SORTIE runs from theseparameter sets (i.e., uncertainty analysis runs) yieldeda distribution of potential model results given the ob-served error in growth and mortality responses to ournitrogen addition experiments. By propagating the sta-tistical uncertainty associated with nitrogen effects onseedlings to SORTIE predictions of relative speciesbasal area, we identified robust predictions of potentialchanges in composition in response to nitrogen avail-ability. We took a conservative approach to assessing

significant differences in species relative basal areasbetween control and nitrogen treatments. We evaluatedsignificant differences with respect to the degree ofoverlap between the central 95% or 90% of control vs.nitrogen treatment in both the baseline and uncertaintyanalysis runs. These 95% and 90% criteria of non-overlap correspond to rough approximations of prob-ability levels of 0.05 and 0.10.

In addition, simple Pearson correlation coefficientswere calculated between variation in each of the alteredparameters in the model and species’ relative abun-dance at different times. Each of the model’s runs rep-resented one sample point in each of the correlations,and 210 such correlations were calculated (7 parame-ters 3 6 species 3 5 time points). The squared cor-relation coefficients (r2) were used to measure the pro-portion of the variance in the relative abundance ofeach species explained by each parameter (Turner etal. 1994).

RESULTS

Seedling gap experiment

Nitrogen additions led to a significant increase inavailability of both ammonium and nitrate in the soilmatrix (F1,96 5 45.89 for ammonium, 49.81 for nitrate,P , 0.001 for both). Ammonium increased from 0.027to 0.080 mg/g resin, and nitrate from 0.088 to 0.490mg/g resin (0 vs. 7.5 g N·m22·yr21 treatments). Thiseffect was consistent across individually and compet-itively grown seedlings (no significant nitrogen 3 den-sity interaction, F1,96 5 2.20 for ammonium, 0.66 fornitrate, P . 0.05 for both). Resin bags trapped morenitrate overall than ammonium, but in both cases, ni-trogen increased the amount of ion extracted from theresin.

For both individually and competitively grown seed-lings, increased soil nitrogen availability led to in-creased growth of only the two fastest growing species(Fig. 1; significant nitrogen 3 species interactions inboth models, F5, 243 5 2.77 and P , 0.05 for individuals,F5, 726 5 4.56, and P , 0.001 for stands). Red mapleshowed significant nitrogen-induced growth enhance-ments both as an individual and in competition, whileyellow birch only showed biomass enhancements whengrown individually and not in mixed-species treat-ments. The other broad-leaved species, sugar maple,and all the conifer species showed no significant growthresponses to nitrogen addition. For competitivelygrown seedlings, there was a significant positive cor-relation (r 5 0.92, P , 0.01) between early seedlinggrowth rates (as assessed by mean biomass in controltreatments) and species’ responsiveness to nitrogen(slope of the regression between biomass and nitrogenaddition level). For individually grown seedlings, therewas a marginally significant positive correlation (r 50.70, P 5 0.12). In contrast to growth, seedling mor-tality in high light was not influenced by nitrogen avail-


FIG. 1. Effects of nitrogen addition (white 5 0, light gray5 2.5, and dark gray 5 7.5 g N·m22·yr21) on final biomassof seedlings after 16 mo of growth (mean 6 1 SE), growneither (a) as individuals or (b) in mixed-species stands. Meansand standard errors were calculated from natural-logarithmtransformed data and then back-transformed (but note logscale). Values above each species represent slope coefficientscalculated using transformed data. Slopes significantly dif-ferent from zero (P , 0.05) are denoted with an asterisk.

FIG. 2. Changes in mean adult relative basalarea ( y-axis) through time generated from SOR-TIE simulations with original control parame-ters (from Pacala et al. 1996) for the six studyspecies.

ability over the course of the two years (no significantnitrogen or nitrogen 3 species effects, F1, 259 5 0.03and F5, 259 5 0.51, P . 0.05 in both cases).

Nitrogen additions increased species’ foliar nitrogenconcentrations, expressed on either an area or massbasis (significant nitrogen effects in ANOVA, F1, 241 555.42 and 33.38, respectively, P , 0.001 for both), andthese responses were consistent across all species (nosignificant nitrogen 3 species interactions, F5, 241 5

1.44 for area and 1.73 for mass, P . 0.05 for both).Overall, nitrogen addition increased foliar nitrogenfrom 16.0 mg/g in control seedlings to 20.5 mg/g inhigh nitrogen seedlings (25% increase). These changesonly led to significant increases in total foliar nitrogencontent (F1,50 . 17.0, P , 0.001) for the two broad-leaved species that grew larger with added nitrogen(yellow birch and red maple; 4.5 and 2.6 times greaterfrom control to high nitrogen, respectively) and for theone conifer species that showed a tendency towardsincreased growth (white pine, 1.9 times greater).

Modeled effects on composition

In all SORTIE control runs (original parametersused, no nitrogen effects), conifers consistently dom-inated the forest community (Fig. 2), with white pinemaking the largest contribution to community com-position at 200 yr (29% of total basal area), with furtherdominance by the late-successional conifers, hemlockand red spruce, developing over the next 300 yr (eachend close to 36%). Yellow birch remained a consistentcomponent of the forest community (10–13% through-out), while both maple species showed steady declinesin relative abundance after an initial peak around 50yr (14% for red maple, 9% for sugar maple).

To examine potential effects of nitrogen availabilityon forest dynamics mediated through seedling regen-eration, SORTIE parameters were adjusted based onthe results of both of our seedling experiments (Table2). As described above, in high-light simulated gapconditions, nitrogen only influenced seedling growthand not mortality, and it was only the fast-growingbroad-leaved species that showed a significant responseto increased nitrogen at early growth stages. Therefore,in the nitrogen SORTIE simulations, the high-lightgrowth parameter (G1) was adjusted for both yellowbirch and red maple (Table 2). In contrast, in the lowlight conditions of the forest understory, nitrogen ad-dition only influenced seedling mortality and notgrowth. All study species except sugar maple were sig-nificantly affected by increased nitrogen to some de-


TABLE 2. Nitrogen-induced changes in SORTIE growth andmortality parameters. Effects are shown as change in thenitrogen treatment relative to control (GrowthN/GrowthC or%MortalityN/%MortalityC), with only changes that werestatistically significant (P , 0.05) included.

Species

High-light growthparameter (G1)

Low N High N

Low-light mortalityparameter (M1)

Low N High N

Yellow birchRed mapleSugar mapleWhite pineRed spruceHemlock

1.361.16

············

1.721.27

············

1.441.90

···2.001.090.93

1.542.56

···2.351.100.03

Notes: Low N corresponds to 2.5 g N·m22·yr21, and highN corresponds to 7.5 g N·m22·yr21. Values greater than 1correspond to a nitrogen-induced increase in the parameter,while values less than 1 correspond to a nitrogen-induceddecrease in the parameter. For red maple, the M2 parameter(mortality sensitivity to carbon balance) was altered instead(see Materials and methods: SORTIE parametrization: Mor-tality for details).

FIG. 3. SORTIE simulations showing effects of nitrogenaddition (white 5 0, light gray 5 2.5, and dark gray 5 7.5g N·m22·yr21) on species’ adult relative basal area at differenttime points in succession (25, 50, 100, 250, and 500 yr). Barheights correspond with the species’ mean relative basal areafrom 40 replicate runs of SORTIE. Error bars correspond withpredicted relative basal areas from the central 95% of thesimulations and represent inherent stochasticity in the modeloriginating from probabilistic submodels in SORTIE. Nitro-gen treatments that led to a significant change in relative basalarea for a given species–time combination are shown withsymbols above the bars: †, no overlap in central 90% of modelprediction; *, no overlap in central 95% of model prediction.

gree (Catovsky and Bazzaz 2002a). Nitrogen additionincreased mortality of white pine and red maple bymore than two-fold, while the nitrogen-induced in-creases in mortality of yellow birch and red spruce wereconsiderably smaller (Table 2). Hemlock was the onlyspecies whose mortality was significantly decreased byincreased nitrogen availability.

Increasing nitrogen availability altered these basicsuccessional dynamics by favoring particular speciesat different times in the course of succession. We firstfocused on effects of nitrogen addition on predictedadult relative basal areas, taking into account only thebounds of the stochastic variation in SORTIE (baselineruns) (Fig. 3). Increased nitrogen availability led toincreased community contribution from yellow birchthroughout all but the very late stages of succession(500 yr). Very early on (25 yr), yellow birch relativebasal area increased from 12% in the control to 18%in low nitrogen and 24% in high nitrogen, and thisincrease persisted throughout the first 250 yr of suc-cession (Fig. 3a–d). At later stages of succession (250and 500 yr), nitrogen availability enhanced communitydominance by the most late-successional species in thesystem, hemlock (from 29% to 34% and 42%) (Fig.3d, e). In contrast to these increases, three mid-suc-cessional species (red maple, white pine, red spruce)all showed declines in their representation within themixed forest community. Nitrogen-induced declines inred maple began relatively early in succession (at 50yr, 14% control vs. 12% low nitrogen and 10% highnitrogen) and, like yellow birch, persisted throughoutall but the very late stages (up to 250 yr) (Fig. 3a–d).Nitrogen led to decreased abundance of white pine be-tween 100 and 500 yr (on average, 22% control vs.18% low nitrogen and 14% high nitrogen), while de-clines for red spruce only occurred at the beginning


FIG. 4. Uncertainty analysis of SORTIE simulations show-ing effects of nitrogen addition (white 5 0, light gray 5 2.5,and dark gray 5 7.5 g N·m22·yr21) on species’ adult relativebasal area at different time points in succession (25, 50, 100,250, and 500 yr). Bar heights correspond with the species’ meanrelative basal area from 40 runs of SORTIE, with parametersgenerated from a distribution based on the error in the experi-mental effects of nitrogen addition. Error bars correspond withpredicted relative basal areas from the central 95% of thesesimulations, representing uncertainty in model predictions aris-ing from observed error in nitrogen treatments. Nitrogen treat-ments that led to a significant change in relative basal area fora given species–time combination are shown with symbols abovethe bars: †, no overlap in central 90% of model prediction; *,no overlap in central 95% of model prediction.

(25 yr) and end (500 yr) of the successional sequence(Fig. 3a, e).

Our uncertainty analysis revealed which of thesecommunity level predictions were robust to observederror in seedling growth and mortality responses tonitrogen (Fig. 4). Here we again assessed significantnitrogen effects on community composition by exam-ining overlap in the central 95% (and 90%) of predic-tions, which in this case incorporated both stochasticvariation in the model and error in seedling growth andmortality responses. Most of the changes in relativebasal area under low nitrogen addition (2.5 gN·m22·yr21) did not remain significant in the uncer-tainty analysis, but a number of the high nitrogenchanges (7.5 g N·m22·yr21) did. Most notable was therobustness of the nitrogen-induced increase in yellowbirch relative basal area early in succession, althoughthe wider error bounds of the uncertainty analysismeant that the increase only extended to 100 yr (Fig.4a–c), rather than 250 yr (compare Figs. 3d and 4d).The significant increase in hemlock relative basal arealater in succession was also robust to experimental er-ror, but only at the very late stages of succession (500yr, Fig. 4e) and not earlier (e.g., 250 yr, Fig. 4d). Forthe other species (red maple, white pine, red spruce),the nitrogen-induced declines in relative basal area thatoccurred in the baseline runs were less robust in theuncertainty analysis. Examining overlap in the central90% of SORTIE predictions, we found significant de-clines in white pine and red maple relative basal areaat mid-successional time points, e.g., 100 and 250 yr(Fig. 4c, d), and red spruce very early in succession(25 yr, Fig. 4a). However, most changes that were sig-nificant for these species in the baseline runs were nolonger significant in the uncertainty analysis.

As well as evaluating the robustness of our modelresults, the uncertainty analysis was used to partitionvariance in species’ relative abundance due to the sep-arate growth and mortality parameters (Fig. 5). Sugarmaple was the only species whose parameters were notaltered in the high nitrogen model runs, and, as a result,changes in model parameters explained relatively little(,30%) of the variance in its relative abundancethrough time (Fig. 5c). For the earlier successional spe-cies (Fig. 5a, b, and d), variation in their individualgrowth/mortality parameters explained much of thevariance in their relative abundance (82% for yellowbirch, 77% for red maple, 53% for white pine; ex-pressed relative to total variance explained in eachcase). For both yellow birch and red maple very earlyin succession, positive growth effects contributed agreater proportion of the variance in relative abundancethan did negative mortality effects (e.g., 22% vs. 7%for yellow birch, 23% vs. 11% for red maple). However,this effect was quickly reversed, particular for red ma-ple, whose growth contributions did not persist beyondthe first 25 yr (Fig. 5b). By the end of the successionalsequence, nitrogen-induced changes in mortality ex-


FIG. 5. Percentage of variance in relative abundance of the study species at different time points (25, 50, 100, 250, and500 yr) explained by changes in model parameters (G 5 species growth function altered, M 5 species mortality functionaltered). Variance values were obtained from the squared Pearson correlation coefficients (r2), and are only shown for thehigh nitrogen SORTIE runs (7.5 g N·m22·yr21).

plained 65–70% of the variance in relative abundancefor both these species. Similarly, for white pine, mor-tality only explained a substantial proportion of vari-ance in relative abundance after 100 yr, paralleling theonset of significant negative impacts of nitrogen onwhite pine relative abundance.

In contrast to this direct link between nitrogen-induced changes in species’ model parameters andtheir relative abundance, the dynamics of the latersuccessional species (red spruce and hemlock) werestrongly determined by changes in the growth andmortality of other species throughout much of suc-cession (Fig. 5e, f). Up until 250 yr, nitrogen-inducedincreases in red spruce mortality explained ,1% ofthe variance in its relative abundance, while decreas-es in hemlock mortality explained ,10% of the var-iance in its relative abundance. However, by 500 yr,when hemlock first showed a robust increase in rel-ative abundance (Fig. 4e), nitrogen-induced decreas-es in hemlock seedling mortality explained .30% ofthe total variance in hemlock relative abundance(Fig. 5f). Similarly, for red spruce, nitrogen effectson mortality also contributed more to variance inrelative abundance by 500 yr, but their contribution

was still restricted to ,20%, and consequently didnot lead to robust declines in red spruce relativeabundance later in succession (Fig. 4e).

DISCUSSION

Species’ responsiveness to increasednitrogen availability

Early seedling growth rate was a good predictor ofspecies’ responsiveness to increased nitrogen avail-ability. When grown both individually and in compe-tition, earlier successional broad-leaved species (yel-low birch and red maple) consistently showed the great-est increases in biomass in response to nitrogen addi-tion, while the most late-successional broad-leavedspecies (sugar maple) and all the coniferous speciesexamined did not grow significantly larger in treat-ments with higher nitrogen availability. We found asignificant correlation between a species’ growth rateduring the first two years of its life and its growthenhancement following nitrogen addition, suggestingthat growth rate, more than leaf habit or even succes-sional position, may be a species’ trait closely tied tonitrogen responsiveness. Of course, all of these traits


(leaf habit, successional position, growth rate) are in-timately related (Reich et al. 1998a), and thus all con-tribute in some way to determining the degree of ni-trogen growth enhancement for a species. Preferencefor ammonium vs. nitrate could not be assessed in thisexperiment, but may be an important driver of seedlingregeneration and community dynamics (Crabtree andBazzaz 1993, McKane et al. 2002). Temperate forestsoils are ammonium dominated, but often show peaksof nitrification (and nitrate concentrations) followingdisturbance (Mladenoff 1987, Bradley 2001).

The emergence of species growth rate as a good pre-dictor of nitrogen responsiveness supports basic eco-logical theory that fast-growing species are more re-sponsive to increases in resource availability (Grime1979, Bazzaz 1996). Species’ with high inherentgrowth rates typically have a suite of traits that givethem the capacity to take up nutrients rapidly from thesoil (Lajtha 1994, Aerts and Chapin 2000) and utilizethese nutrients effectively to increase carbon uptakeand plant growth (Grime et al. 1997, Reich et al.1998b), while slow-growing species place greater em-phasis on high nutrient retention and high nutrient useefficiency, at the expense of a greater potential for re-source capture (Chapin 1980, Aerts and Chapin 2000).This relationship was particularly strong for seedlingsgrown in a competition treatment where nitrogen-in-duced fast early growth allowed certain seedlings togrow ahead of their competitors and thus experiencereduced competition for light (Berntson and Wayne2000). Yellow birch showed lower responsiveness toincreased nitrogen availability under competition thanas individuals, as its small initial seedling size oftenplaced it subordinate to red maple in mixed-speciesstands. Plant competition often changes predictionsabout species’ responses to global environmentalchange (Catovsky and Bazzaz 2002b).

What are the mechanisms underlying these species-specific responses? All species had greater foliar ni-trogen concentrations with increasing nitrogen avail-ability (except for red spruce), but the more responsivespecies were able to take up more nitrogen from thesoil on a whole-plant basis. Species whose growth wasnot significantly affected by nitrogen amendment (co-nifers particularly) apparently were not able to utilizethe additional nitrogen in their leaves for photosyn-thesis. The photosynthetic capacity of evergreen co-niferous species has been shown to be much less re-sponsive to foliar nitrogen concentrations than that ofdeciduous broad-leaved species (Reich et al. 1995, Shi-nano et al. 2001), although the physiological mecha-nisms for these differences are still equivocal (Becker2000). Conifers might be less responsive to increasesin foliar nitrogen than broad-leaved species, if theywere to allocate a greater proportion of the additionalnitrogen to structural rather than photosynthetic pro-teins (Bazzaz 1997). Alternatively, leaf structure mightimpose a greater stomatal limitation on photosynthesis

in conifers than in broad-leaved species (Sharkey1985), so that conifers have a lower capacity to utilizeadditional nitrogen for photosynthesis. More detailedleaf-level physiological studies are needed to ascertainwhat determines a species’ capacity to make use ofhigher foliar nitrogen concentrations.

Scaling up seedling responses to nitrogen

Our understanding of controls on forest dynamics isconstrained by the short-term and small-scale nature ofmost ecological experiments (individual seedlingsgrown in pots for up to three growing seasons) relativeto the appropriate scale of interest (whole-ecosystemdynamics over many generations) (Bazzaz et al. 1996).As a result, we need to utilize different ways to ex-trapolate from the scale of our experimental studies tohigher levels of organization (Ehleringer and Field1993). In the current paper, we integrated the resultsof two field-based experimental manipulations into anempirically based forest dynamics model (SORTIE) toexamine how the influence of nitrogen availability onseedling growth and mortality might affect our under-standing of structure and dynamics of temperate forestsin eastern North America. Current forest models areprimarily driven by effects on and responses to lightavailability (Pacala et al. 1996). We now need to de-termine how such models can be applied to a range offorests within the same region, especially as soil fer-tility changes.

Our experimental work showed that, in the high lightconditions of a simulated forest gap, increased nitrogenfavored growth of fast-growing broad-leaved species(yellow birch and red maple), while in the low lightconditions of forest understory, nitrogen promoted sur-vival of the most shade-tolerant conifer species (hem-lock), and increased mortality of a number of less tol-erant species (red maple and white pine, in particular)(Table 2). Incorporating these changes into SORTIErevealed that seedling responses to nitrogen availabilitycould act as an important driver of community dynam-ics in these mixed forests, potentially altering succes-sional rates of change.

Nitrogen-induced changes at the community levelarose from the combined effects of nitrogen on twodistinct components of seedling regeneration: growthin high light and mortality in low light (Table 2 andFig. 5). In some cases, the results of nitrogen on onecomponent of seedling regeneration could be directlyrelated to the SORTIE outcomes. For example, the de-crease in white pine from 100 yr onwards could beattributed primarily to nitrogen-induced increases inseedling mortality in the understory for this species. Inother cases, the resultant dynamics were a more com-plex combination of nitrogen effects on both growthand mortality. For both yellow birch and red maple,their changes in relative abundance through successionresulted from a trade-off between the positive effectsof nitrogen on high-light growth and the negative ef-


fects of nitrogen on low-light survival. Yellow birchbenefited from increased nitrogen availability on thewhole, because the positive effects predominated earlyin succession and persisted for the first 100 yr, whilethe negative effects did not come into play until laterin succession when other species exhibited even stron-ger negative effects of nitrogen on survival. By con-trast, nitrogen reduced red maple abundance on thewhole, because the negative effects of nitrogen on low-light survival outweighed the positive effects on high-light growth at all but the very earliest stages of suc-cession. A third class of scaling between seedling re-generation and community dynamics was found for thelate-successional conifer species, whose dynamicswere dependent on nitrogen effects not only on theregeneration of their own seedlings, but also the seed-lings of other species. For example, the increased rel-ative abundance of hemlock later in succession aroseboth from increased low-light survival of its own seed-lings, but also from decreased survival of seedlings ofwhite pine, red maple, and particularly yellow birch.

Through its effects on seedling regeneration, highernitrogen availability led to exaggerated successionaldynamics in mixed temperate forests. The nitrogen-enhanced growth of faster growing species (yellowbirch) led to their increased dominance earlier in suc-cession, while the nitrogen-enhanced survivorship oflater successional species (particularly hemlock) en-abled them to maintain a more persistent seedling bankin the model, and enabled hemlock to increase its dom-inance in late-successional forests. This main findingwas robust to an uncertainty analysis that incorporatedexperimentally derived error into the seedling/saplinggrowth and mortality submodels. In contrast, the iden-tity of the species replaced by yellow birch and hem-lock was less certain and more sensitive to uncertaintyin our parameter values. Red maple and white pine arethe most likely candidates to decline in abundance withhigher nitrogen availability, but this outcome is likelyto be site dependent.

The SORTIE model provided a useful way to ex-amine the potential for seedling nitrogen responses toscale up and influence forest successional dynamics.However, the results should not be seen as definitivepredictions of overall forest responses to enhanced ni-trogen availability, because soil nitrogen will affectother components of forest dynamics beyond the seed-ling/sapling stage, e.g., growth, survivorship, and re-production of mature trees (Mitchell and Chandler1939, Magill et al. 1997). Even as a tool for examiningthe larger scale significance of nitrogen effects on seed-ling regeneration, this modeling approach makes cer-tain assumptions: (1) that the changes observed forseedlings in the first two years are carried throughoutthe seedling/sapling stage, i.e., until reaching 10 cmdbh when the adult growth function takes over, and (2)that nitrogen-induced changes will apply equally toseedlings at all points in succession, i.e., whether 25

or 250 yr after a major disturbance. To establish wheth-er these are reasonable assumptions or not, we need todetermine the degree to which seedlings acclimate tonovel environmental conditions (e.g., Bazzaz et al.1993), and the extent to which nitrogen availabilitychanges through succession (e.g., Vitousek and Reiners1975).

Implications for future forest dynamics

Increasing human impacts on natural ecosystems hasheightened the need to understand controls on ecosys-tem structure and function (Vitousek et al. 1997b). Amultiple resource perspective is useful in this regard,as many novel perturbations involve changes in seed-lings’ resource environment (Field et al. 1992, Bazzazand Catovsky 2002). If we can understand how speciesbehave along a suite of resource axes, we should beable to predict patterns of forest dynamics under arange of future scenarios. For example, nitrogen de-position in temperate regions has emerged as one ofthe most dramatic human-induced changes since in-dustrialization (Vitousek et al. 1997a). Natural eco-systems in all these temperate regions are commonlynitrogen limited (Vitousek and Howarth 1991), and ni-trogen deposition might exert profound effects on theirstructure and function (Magill et al. 1997, Aber et al.1998). Our SORTIE results have some bearing on pre-dicting impacts of nitrogen deposition on mixed tem-perate forests, because using the model we were ableto assess potential consequences of nitrogen-inducedchanges in seedling regeneration for long-term forestdynamics. Our findings suggest that, through effectson seedling growth and survival, increased nitrogenloading in the future will further favor growth of earliersuccessional species following natural disturbanceevents, while later successional species (particularlyhemlock) will be able to maintain a more persistentseedling bank, and thus remain a dominant componentof late-successional forests.

Increased nitrogen loading in the future will mostlikely change temperate forest community structure anddynamics, but the exact nature of any change will bedetermined by: (1) long-term ecosystem consequencesof nitrogen deposition, and (2) interactions with otherhuman impacts on forests. Our relatively simple mod-eling analysis does not account for a suite of othernitrogen-induced ecosystem-level changes that will allaffect long-term forest responses to nitrogen deposi-tion, e.g., degree of nitrogen retention within forests(see nitrogen saturation hypotheses; Aber et al. 1998,Emmett et al. 1998), sequestration of nitrogen withinsoil organic matter (Gundersen et al. 1998), timing andextent of vegetation uptake of nitrogen from the soil(Boxman et al. 1998), and leaching of base cations (Ca,Mg, and K) (Likens et al. 1996). In addition, humanactivities will exert a range of other perturbations ontemperate forests that will interact with these nitrogeneffects (Aber et al. 2001), such as (1) direct manage-


ment of forests and utilization of wood products (Irlandet al. 2001), (2) introduction of invasive species, suchas the hemlock woolly adelgid (Adelges tsugae An-nand) (McManus et al. 2000), and (3) altered atmo-spheric composition, such as elevated CO2 (Bolker etal. 1995). Incorporating modified seedling parametersinto the SORTIE model provided a useful first indi-cation of how nitrogen availability could potentiallyinfluence long-term forest dynamics through its im-pacts on seedling regeneration. Future attempts to pre-dict the impacts of environmental change on foreststructure and function should consider both these lon-ger term ecosystem-level impacts, as well the effectson individual tree performance beyond its first fewyears as a seedling (e.g., Bazzaz et al. 1993).

ACKNOWLEDGMENTS

We thank Henry Schumacher for much assistance with theexperimental work, and Joe LaCasse and David Hembry forhelp with setting up and harvesting the experiment. RichardCobb, Steve Currie, and David Orwig at Harvard Forest pro-vided tremendous help with the resin bag analyses, while MattKizlinski at Harvard Forest provided generous assistance withthe foliar nitrogen analyses. We are very grateful to GuntramBauer, Christine Muth, Eric Macklin, Amity Wilczek, KristinaStinson, and David McGuire, who all contributed substan-tially to the design and interpretation of the experiment. Wethank John Genet for completing the SORTIE runs. This workwas supported by a NASA Earth System Science GraduateFellowship, a Student Dissertation Grant from the Departmentof Organismic and Evolutionary Biology, and a Mellon Foun-dation Grant from Harvard Forest (all to S. Catovsky), andby the Harvard Forest Long-Term Ecological Research Pro-gram (NSF Grant DEB 9411795 to F. A. Bazzaz) and Mich-igan Agricultural Experiment Station (Project 1693, Multi-State Research Project NRSP-3 to R. K. Kobe).

LITERATURE CITED

Aber, J., W. McDowell, K. Nadelhoffer, A. Magill, G. Bernt-son, M. Kamakea, S. McNulty, L. Rustad, and I. Fernandez.1998. Nitrogen saturation in temperate forest ecosystems.BioScience 48:921–934.

Aber, J. D., R. P. Neilson, S. McNulty, J. M. Lenihan, D.Bachelet, and R. J. Drapek. 2001. Forest processes andglobal environmental change: predicting the effects of in-dividual and multiple stressors. BioScience 51:735–751.

Aerts, R., and F. S. Chapin, III. 2000. The mineral nutritionof wild plants revisited: a re-evaluation of processes andpatterns. Advances in Ecological Research 30:1–67.

Bazzaz, F. A. 1996. Plants in changing environments: linkingphysiological, population, and community ecology. Cam-bridge University Press, Cambridge, UK.

Bazzaz, F. A. 1997. Allocation of resources in plants: state-of-the-science and critical questions. Pages 1–37 in F. A.Bazzaz and J. Grace, editors. Plant resource allocation. Ac-ademic Press, San Diego, California, USA.

Bazzaz, F. A., S. L. Bassow, G. M. Berntson, and S. C. Thom-as. 1996. Elevated CO2 and terrestrial vegetation: impli-cations for and beyond the global carbon budget. Pages 43–76 in B. Walker and W. Steffen, editors. Global change andterrestrial ecosystems. Cambridge University Press, Cam-bridge, UK.

Bazzaz, F. A., and S. Catovsky. 2002. Plants—from cells toecosystems: impacts of global environmental change. Pages94–111 in H. A. Mooney and J. G. Canadell, editors. En-cyclopedia of global environmental change. Volume 2.John Wiley and Sons, Chichester, UK.

Bazzaz, F. A., S. L. Miao, and P. M. Wayne. 1993. CO2-induced growth enhancements of co-occurring tree speciesdecline at different rates. Oecologia 96:478–482.

Bazzaz, F. A., and P. M. Wayne. 1994. Coping with envi-ronmental heterogeneity: the physiological ecology of treeseedling regeneration across the gap-understory continu-um. Pages 349–390 in M. M. Caldwell and R. W. Pearcy,editors. Exploitation of environmental heterogeneity byplants: ecophysiological processes above- and below-ground. Academic Press, San Diego, California, USA.

Becker, P. 2000. Competition in the regeneration niche be-tween conifers and angiosperms: Bond’s slow seedling hy-pothesis. Functional Ecology 14:401–412.

Berntson, G. M., and P. M. Wayne. 2000. Characterizing thesize dependence of resource acquisition within crowdedplant populations. Ecology 81:1072–1085.

Binkley, D., and P. Vitousek. 1989. Soil nutrient availability.Pages 75–96 in R. W. Pearcy, J. R. Ehleringer, H. A. Moo-ney, and P. W. Rundel, editors. Plant physiological ecology:field methods and instrumentation. Chapman and Hall,London, UK.

Bolker, B. M., S. W. Pacala, F. A. Bazzaz, C. D. Canham,and S. A. Levin. 1995. Species diversity and ecosystemresponse to carbon dioxide fertilization: conclusions froma temperate forest model. Global Change Biology 1:373–381.

Bond, W. J. 1989. The tortoise and the hare: ecology ofangiosperm dominance and gymnosperm persistence. Bi-ological Journal of the Linnean Society 36:227–249.

Boxman, A. W., K. Blanck, T.-E. Brandrud, B. A. Emmett,P. Gundersen, R. F. Hogervorst, O. J. Kjønaas, H. Persson,and V. Timmerman. 1998. Vegetation and soil biota re-sponse to experimentally-changed nitrogen inputs in co-niferous forest ecosystems of the NITREX project. ForestEcology and Management 101:65–79.

Bradley, R. L. 2001. An alternative explanation for the post-disturbance NO3

2 flush in some forest ecosystems. EcologyLetters 4:412–416.

Burns, R. M., and B. H. Honkala, editors. 1990. Silvics ofNorth America. Volume 1. Conifers. United States De-partment of Agriculture, Washington, D.C., USA.

Canham, C. D. 1989. Different responses to gaps amongshade-tolerant tree species. Ecology 70:548–550.

Canham, C. D., A. R. Berkowitz, V. R. Kelly, G. M. Lovett,S. V. Ollinger, and J. Schnurr. 1996. Biomass allocationand multiple resource limitation in tree seedlings. CanadianJournal of Forest Research 26:1521–1530.

Canham, C. D., A. C. Finzi, S. W. Pacala, and D. H. Burbank.1994. Causes and consequences of resource heterogeneityin forests: interspecific variation in light transmission bycanopy trees. Canadian Journal of Forest Research 24:337–349.

Canham, C. D., and O. L. Loucks. 1984. Catastrophic wind-throw in the presettlement forests of Wisconsin. Ecology65:803–809.

Caspersen, J. P., and R. K. Kobe. 2001. Interspecific variationin sapling mortality in relation to growth and soil moisture.Oikos 92:160–168.

Catovsky, S., and F. A. Bazzaz. 2000. The role of resourceinteractions and seedling regeneration in maintaining a pos-itive feedback in hemlock stands. Journal of Ecology 88:100–112.

Catovsky, S., and F. A. Bazzaz. 2002a. Nitrogen availabilityinfluences regeneration of coniferous and broad-leaved treespecies in the understory seedling bank. Ecological Ap-plications 12:1056–1070.

Catovsky, S., and F. A. Bazzaz. 2002b. Plant competition inan elevated CO2 world. Pages 471–481 in H. A. Mooneyand J. G. Canadell, editors. Encyclopedia of global envi-


ronmental change. Volume 2. John Wiley and Sons, Chich-ester, UK.

Catovsky, S., and F. A. Bazzaz. In press. Feedbacks betweencanopy composition and seedling regeneration in mixedconifer broad-leaved forests. Oikos.

Chapin, F. S., III. 1980. The mineral nutrition of wild plants.Annual Review of Ecology and Systematics 11:233–260.

Crabtree, R. C., and F. A. Bazzaz. 1993. Seedling responseof four birch species to simulated nitrogen deposition: am-monium vs. nitrate. Ecological Applications 3:315–321.

de Vries, P. G. 1986. Sampling theory for forest inventory:a teach-yourself course. Springer-Verlag, New York, NewYork, USA.

Ehleringer, J. R., and C. B. Field, editors. 1993. Scalingphysiological processes: leaf to globe. Academic Press, SanDiego, California, USA.

Emmett, B. A., D. Boxman, M. Bredemeier, P. Gundersen,O. J. Kjønaas, F. Moldan, P. Schleppi, A. Tietema, and R.F. Wright. 1998. Predicting the effects of atmospheric ni-trogen deposition in conifer stands: evidence from the NI-TREX ecosystem-scale experiments. Ecosystems 1:352–360.

Field, C. B., F. S. Chapin, III, P. A. Matson, and H. A. Moo-ney. 1992. Responses of terrestrial ecosystems to thechanging atmosphere: a resource-based approach. AnnualReview of Ecology and Systematics 23:201–235.

Finzi, A. C., and C. D. Canham. 2000. Sapling growth inresponse to light and nitrogen availability in a southernNew England forest. Forest Ecology and Management 131:153–165.

Foster, D. R., G. Motzkin, and B. Slater. 1998. Land-usehistory as long-term broad-scale disturbance: regional for-est dynamics in central New England. Ecosystems 1:96–119.

Foster, D. R., D. A. Orwig, and J. S. McLachlan. 1996. Eco-logical and conservation insights from reconstructive stud-ies of temperate old-growth forests. Trends in Ecology andEvolution 11:419–424.

Grime, J. P. 1979. Plant strategies and vegetation processes.John Wiley and Sons, Chichester, UK.

Grime, J. P., K. Thompson, R. Hunt, J. G. Hodgson, J. H. C.Cornelissen, I. H. Rorison, G. A. F. Hendry, T. W. Ash-enden, A. P. Askew, S. R. Band, R. E. Booth, C. C. Bossard,et al. 1997. Integrated screening validates primary axes ofspecialisation in plants. Oikos 79:259–281.

Grubb, P. J. 1994. Root competition in soils of different fer-tility: a paradox resolved? Phytocoenologia 24:495–505.

Gundersen, P., B. A. Emmett, O. J. Kjønaas, C. J. Koopmans,and A. Tietema. 1998. Impact of nitrogen deposition onnitrogen cycling in forests: a synthesis of NITREX data.Forest Ecology and Management 101:37–55.

Irland, L. C., D. Adams, R. Alig, C. J. Betz, C.-C. Chen, M.Hutchins, B. A. McCarl, K. Skog, and B. L. Sohngen. 2001.Assessing socioeconomic impacts of climate change onU.S. forests, wood-product markets, and forest recreation.BioScience 51:753–764.

Kobe, R. K. 1996. Intraspecific variation in sapling mortalityand growth predicts geographic variation in forest com-position. Ecological Monographs 66:181–201.

Kobe, R. K., S. W. Pacala, J. A. Silander, Jr., and C. D.Canham. 1995. Juvenile tree survivorship as a componentof shade tolerance. Ecological Applications 5:517–532.

Lajtha, K. 1994. Nutrient uptake in eastern deciduous treeseedlings. Plant and Soil 160:193–199.

Likens, G. E., C. T. Driscoll, and D. C. Buso. 1996. Long-term effects of acid rain: response and recovery of a forestecosystem. Science 272:244–246.

Magill, A. H., J. D. Aber, J. J. Hendricks, R. D. Bowden, J.M. Melillo, and P. A. Steudler. 1997. Biogeochemical re-

sponse of forest ecosystems to simulated chronic nitrogendeposition. Ecological Applications 7:402–415.

Marks, P. L., and S. Gardescu. 1998. A case study of sugarmaple (Acer saccharum) as a forest seedling bank species.Journal of the Torrey Botanical Society 125:287–296.

McKane, R. B., L. C. Johnson, G. R. Shaver, K. J. Nadel-hoffer, E. B. Rastetter, B. Fry, A. E. Giblin, K. Kielland,B. L. Kwiatkowski, J. A. Laundre, and G. Murray. 2002.Resource-based niches provide a basis for plant speciesdiversity and dominance in arctic tundra. Nature 415:68–71.

McManus, K. A., K. S. Shields, and D. R. Souto, editors.2000. Proceedings: symposium on sustainable manage-ment of hemlock ecosystems in eastern North America.USDA Forest Service, Newtown Square, Pennsylvania,USA.

Mitchell, H. L., and R. F. Chandler, Jr. 1939. The nitrogennutrition and growth of certain deciduous trees of north-eastern United States. Black Rock Forest Bulletin 11:1–94.

Mladenoff, D. J. 1987. Dynamics of nitrogen mineralizationand nitrification in hemlock and hardwood treefall gaps.Ecology 68:1171–1180.

Munger, J. W., S. M. Fan, P. S. Bakwin, M. L. Goulden, A.H. Goldstein, A. S. Colman, and S. C. Wofsey. 1998. Re-gional budgets for nitrogen oxides from continental sourc-es: variations of rates for oxidation and deposition withseason and distance from source regions. Journal of Geo-physical Research–Atmospheres 103(D7):8355–8368.

Oliver, C. D., and E. P. Stephens. 1977. Reconstruction of amixed-species forest in central New England. Ecology 58:562–572.

Ollinger, S. V., J. D. Aber, G. M. Lovett, S. E. Millham, R.G. Lathrop, and J. M. Ellis. 1993. A spatial model of at-mospheric deposition for the northeastern U.S. EcologicalApplications 3:459–472.

Pacala, S. W., C. D. Canham, J. Saponara, J. A. Silander, Jr.,R. K. Kobe, and E. Ribbens. 1996. Forest models definedby field measurements: estimation, error analysis and dy-namics. Ecological Monographs 66:1–43.

Pacala, S. W., C. D. Canham, J. A. Silander, Jr., and R. K.Kobe. 1994. Sapling growth as a function of resources ina north temperate forest. Canadian Journal of Forest Re-search 24:2174–2183.

Peet, R. K., and N. L. Christensen. 1987. Competition andtree death. BioScience 37:586–595.

Peterken, G. F. 1996. Natural woodland. Cambridge Univer-sity Press, Cambridge, UK.

Reich, P. B., B. D. Kloeppel, D. S. Ellsworth, and M. B.Walters. 1995. Different photosynthesis-nitrogen relationsin deciduous hardwood and evergreen coniferous tree spe-cies. Oecologia 104:24–30.

Reich, P. B., M. G. Tjoelker, M. B. Walters, D. Vanderklein,and C. Buschena. 1998a. Close association of RGR, leafand root morphology, seed mass and shade tolerance inseedlings of nine boreal tree species grown in high and lowlight. Functional Ecology 12:327–338.

Reich, P. B., M. B. Walters, M. G. Tjoelker, D. Vanderklein,and C. Buschena. 1998b. Photosynthesis and respirationrates depend on leaf and root morphology and nitrogenconcentration in nine boreal tree species differing in rel-ative growth rate. Functional Ecology 12:395–405.

Ribbens, E., J. A. Silander, Jr., and S. W. Pacala. 1994. Seed-ling recruitment in forests: calibrating models to predictpatterns of tree seedling dispersion. Ecology 75:1794–1806.

Sharkey, T. D. 1985. Photosynthesis in intact leaves of C3

plants: physics, physiology and rate limitation. BotanicalReview 51:53–105.

Shinano, T., M. Osaki, and M. Kato. 2001. Differences in


nitrogen economy of temperate trees. Tree Physiology 21:617–624.

Shugart, H. H. 1998. Terrestrial ecosystems in changing en-vironments. Cambridge University Press, Cambridge, UK.

Sokal, R. R., and F. J. Rohlf. 1995. Biometry. W. H. Freemanand Company, New York, New York, USA.

ter Mikaelian, M. T., and M. D. Korzukhin. 1997. Biomassequations for sixty-five North American tree species. ForestEcology and Management 97:1–24.

Turner, M. G., Y. Wu, L. L. Wallace, W. H. Romme, and A.Brenkert. 1994. Simulating winter interactions among un-gulates, vegetation, and fire in northern Yellowstone Park.Ecological Applications 4:472–496.

Vitousek, P. M., J. D. Aber, R. W. Howarth, G. E. Likens, P.A. Matson, D. W. Schindler, W. H. Schlesinger, and D. G.Tilman. 1997a. Human alteration of the global nitrogen

cycle: sources and consequences. Ecological Applications7:737–750.

Vitousek, P. M., and R. W. Howarth. 1991. Nitrogen limi-tation on land and in the sea: how can it occur? Biogeo-chemistry 13:87–115.

Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M.Melillo. 1997b. Human domination of Earth’s ecosystems.Science 277:494–499.

Vitousek, P. M., and W. A. Reiners. 1975. Ecosystem suc-cession and nutrient retention: a hypothesis. BioScience 25:376–381.

Walters, M. B., and P. B. Reich. 1997. Growth of Acer sac-charum seedlings in deeply shaded understories of northernWisconsin: effects of nitrogen and water availability. Ca-nadian Journal of Forest Research 27:237–247.

Westveld, M. 1956. Natural forest vegetation zones of NewEngland. Journal of Forestry 54:332–338.

nitrogen-induced changes in seedling regeneration …kobe/docs/catovsky et al 2002 ecol appl.pdf ·...

Documents